JP2005070430A - Speech output device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech output device and a speech output method which can output a high-understandability (easy-to-listen) speech regardless of the contents of an output word. <P>SOLUTION: This device comprises a speech DB 1 wherein familiarity information showing how familiar each word or word string is recorded and a sound pressure adjustment part 3 which adjusts a sound pressure level in units of a word or word string according to familiarity information read out of the speech DB 1 by a reproduction part 2 together with speech data. A word etc., whose familiarity is low is corrected to larger sound pressure and then even when the speech of contents of low word familiarity such as a place name which is unfamiliar is output, it is output with sound pressure larger than a word of high familiarity, so that even the word of low familiarity becomes easy to hear. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to an audio output apparatus and method, and is particularly suitable for use in an apparatus and method for correcting audio so that audio output from an in-vehicle device can be easily heard by a user in a vehicle interior.

  In recent years, there has been an increasing demand for voice output in the vehicle interior, such as the voice of the other party using a hands-free device, the Web information received by the information communication device, and the read-out voice of the e-mail, as well as the guide voice of the navigation device. As the audio output method, a recording / reproduction method for reproducing sound recorded in advance on a medium such as a DVD (Digital Versatile Disk) or a hard disk, and an audio waveform is constructed and reproduced based on given character information. There is a TTS (Text-to-Speech) method.

  The voice output device based on the latter TTS method is roughly divided into a language processing unit for adding reading and accents to given character information by dictionary data for text analysis, and dictionary data of waveform / phoneme segments. It consists of two processing units, a speech synthesis unit that generates speech.

Conventionally, a speech intelligibility improving system based on loudness compensation has been provided for speech output. This system makes it possible to hear the output sound more clearly even in noise by appropriately adjusting the sound pressure level of the output sound according to the level of ambient noise etc. input from the microphone. (For example, see Patent Document 1).
Japanese Patent Laid-Open No. 11-166835

  A conventional speech correction apparatus represented by this intelligibility improvement system corrects the sound pressure of speech based on physical quantities based on ambient noise environments such as noise level and vehicle speed signal. However, even if the output sound is corrected based on these physical quantities so as to be clear, since it is human beings to hear it, not all words and sentences can be understood to the same extent. This is because even if the words have the same sound pressure level, the intelligibility changes depending on the familiarity (degree of familiarity) of the words.

  FIG. 5 is a characteristic diagram of test results showing the relationship among word familiarity, word intelligibility, and sound pressure. This characteristic diagram shows how the word intelligibility changes when the sound pressure level is changed, and also shows how the word intelligibility changes depending on the familiarity of the word being listened to. is there. As is clear from this characteristic diagram, even with the same sound pressure level, the word with higher intimacy has a higher intelligibility, and the word with lower intimacy has a lower intelligibility.

  Thus, even if the sound pressure level of the output sound is corrected based on the physical quantity, the ease of hearing varies depending on the content of the output sound. For this reason, for example, even when the navigation device is running, there is a problem that it becomes more difficult to hear as the place name is unknown.

As an information processing apparatus that takes into account the familiarity of words, there is a kana-kanji conversion apparatus that displays kanjis of conversion candidates arranged side by side in descending order of familiarity (see, for example, Patent Document 2). There is also a pattern recognition device that searches for a word having a higher intimacy and outputs it as a recognition result when there are a plurality of words that express the same concept with respect to the input pattern sequence (for example, Patent Documents). 3).
JP 2001-216295 A JP 2002-162991 A

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a speech with high intelligibility (easy to hear) regardless of the content of the output word. To do.

  In order to solve the above-described problems, in the audio output device of the present invention, information on familiarity indicating how familiar a plurality of words or word strings are is prepared, and the familiarity information is included in the familiarity information. Based on this, the sound pressure level is adjusted in units of words or word strings with respect to the output target speech.

  According to the present invention configured as described above, for example, even in the case of outputting a voice having a low word familiarity such as a land name that cannot be heard, it is output with a larger sound pressure than a word having a high familiarity. By doing so, the intelligibility of the word can be increased. As a result, even when the output voice is composed of words or word strings with low intimacy, or when words with high intimacy or words with low intimacy are mixed in the output voice , You will always be able to provide a high degree of intelligibility.

(First embodiment)
DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, a first embodiment of the invention will be described with reference to the drawings. In the first embodiment, the present invention is applied to an audio output device of a recording / reproducing system. FIG. 1 is a diagram illustrating a configuration example of a main part of an audio output device according to the first embodiment. As shown in FIG. 1, the audio output device according to the present embodiment includes an audio database (DB) 1, a reproduction unit 2, a sound pressure adjustment unit 3, and a volume 4.

  The audio DB 1 is configured by recording waveform-encoded audio data on a medium such as a DVD or a hard disk. In the voice DB 1, voice data to be output is recorded in units of words or word strings. The word string refers to a combination of a plurality of words that are likely to be used simultaneously, an idiom composed of a plurality of words, or a simple sentence. Hereinafter, “words or word strings” are collectively referred to as “words”.

  For example, when the voice output device of this embodiment is applied to a navigation device, a series of guide voices such as “Follow XX direction, traffic jam.” Each is recorded as a separate voice pattern. When reproducing a series of guide voices, a plurality of divided voice patterns are sequentially read out and output.

  The voice DB 1 is further recorded with information on familiarity indicating how familiar each voice pattern is recorded in units of words or the like. This familiarity information is a numerical value (for example, 1.0 to 7.0) indicating how many word familiarities are in units of words or the like. As described above, the voice DB 1 constitutes the familiarity information storage unit of the present invention.

  The reproduction unit 2 reproduces audio data and familiarity information from the audio DB 1. The reproduction unit 2 arbitrarily selects and reads out a plurality of sound patterns from the sound DB 1 (selects and reads a tag at a data position where a sound pattern corresponding to the content to be uttered is recorded), thereby allowing arbitrary contents to be read. It is possible to reproduce the guide voice. At this time, the familiarity information stored corresponding to the read voice patterns is also read.

  The sound pressure adjustment unit 3 controls the volume 4 based on the familiarity information read from the sound DB 1 by the reproduction unit 2, thereby similarly reproducing each sound pattern (each word read from the sound DB 1 by the reproduction unit 2. Etc.) is adjusted. Specifically, the volume value is corrected so as to increase as the voice pattern has a lower familiarity.

  For example, in the case of outputting a guide voice such as “This is the direction of XX ahead, traffic jam”, if the part of XX representing the place name is a word with a high familiarity such as “Koriyama”, the volume value No adjustment is made. On the other hand, in the case where the part of XX representing the place name is a word having a low familiarity such as “Daisen” or “Daizuki”, an adjustment is made to increase the volume value.

  It is preferable that the volume value is increased depending on the word familiarity value. For example, in the characteristic diagram of FIG. 5, assuming that the word intelligibility is 80%, and assuming that the original sound pressure level is about 20 dB, the speech pattern having a word familiarity of 7.0 to 5.5 No adjustment of the sound pressure level is required. On the other hand, a sound pattern having a word familiarity of 5.5 to 4.0 increases the sound pressure level by about 5 dB. In addition, a sound pattern with a word familiarity of 4.0 to 2.5 increases the sound pressure level by about 15 dB, and a sound pattern with a word familiarity of 2.5 to 1.0 increases the sound pressure level by about 20 dB.

  As described above in detail, according to the first embodiment, familiarity information is added and stored in units of words or the like to the voice DB 1 in which each word or the like to be output is recorded, and the voice is recorded. Since the sound pressure level of each word or the like to be output is adjusted as appropriate based on the familiarity information read together with the information, even when outputting a sound whose contents cannot be heard normally, the degree of understanding is high ( (Easy to hear) can be provided. Therefore, for example, even when a navigation device to which the audio output device of the present embodiment is applied is used in an unfamiliar land, the guide audio is always easy to hear.

  In the first embodiment, an example in which a word having the highest word familiarity is set as a reference sound pressure level and a sound pressure level of a word having a lower word familiarity is corrected to a greater value has been described. The invention is not limited to this. For example, the word having the lowest word familiarity may be used as the reference sound pressure level, and the sound pressure level of a word having a higher word familiarity may be corrected to be smaller. In addition, a word having a medium word familiarity is set as a reference sound pressure level, a sound pressure level of a word having a word closeness higher than the reference is corrected to be small, and a sound pressure level of a word having a word closeness lower than the reference is corrected. It is also possible to make the intelligibility of all the words equal to about the same level by correcting.

  In the first embodiment, the example in which the intelligibility of all words and the like is made equal by adjusting the sound pressure level according to the word familiarity has been described. It is not limited. For example, the word intelligibility may be made larger than a predetermined value by adjusting the sound pressure, and it is not always necessary to make the intelligibility of all the words equal.

  In the first embodiment, the example in which the sound pressure level of the sound is adjusted based on the familiarity information has been described. However, in addition to or instead of this, a word whose word familiarity is lower than a predetermined value Etc. may be repeated two or more times. For example, when outputting a guide voice such as “This destination salt direction is congested”, the sound pressure adjustment unit 3 adjusts the sound pressure level to increase at the “difference salt” portion, "It's a traffic jam. It's in the direction of the difference salt." This repetitive playback control can be performed in the playback unit 2. In this way, it is possible to increase the intelligibility by repeatedly reproducing even a word that is not familiar to the user.

  Further, in addition to or instead of adjusting the sound pressure level of sound based on the familiarity information, the sound reproduction speed may be adjusted. For example, in the case of outputting a guide voice such as “This destination salt is in a traffic jam”, the sound pressure adjustment unit 3 adjusts the sound pressure level to increase at the “difference salt” portion, Control the part to play at a slower speed than the other parts. The playback speed can also be controlled by the playback unit 2. In this way, even for words that are not well-known, the intelligibility can be increased by slowing the playback speed.

  Further, in addition to or instead of adjusting the sound pressure level of the voice based on the familiarity information, a word having a word familiarity lower than a predetermined value may be displayed on the screen. This display control can be performed using a display controller (not shown) (for example, in the case of a navigation device, one that is normally provided for displaying a map image or the like on the display device). In this way, even a word or the like that is less familiar is displayed on the screen so that it can be visually confirmed, thereby increasing the degree of understanding.

(Second Embodiment)
Next, a second embodiment of the present invention will be described with reference to the drawings. In the second embodiment, the present invention is applied to a TTS audio output apparatus. FIG. 2 is a diagram illustrating a configuration example of a main part of an audio output device according to the second embodiment. As shown in FIG. 2, the audio output device of the present embodiment includes a text generation unit 11, a TTS engine 12, a sound pressure adjustment unit 13, and a volume 14.

  The text generation unit 11 generates text information that represents the content of audio to be output as a character string. The text generation unit 11 may manually generate text information of an arbitrary character string by a user operating a keyboard (not shown), or the controller may generate text information of an arbitrary character string according to a predetermined rule. May be generated automatically.

  The TTS engine 12 includes a language processing unit 15, a text analysis dictionary 16, a speech synthesis unit 17, and a phoneme unit dictionary 18. The text analysis dictionary 16 is a text analysis dictionary database that stores text information composed of various words and the like, and phoneme information and prosodic information added to the words in association with each other.

  Further, in the text analysis dictionary 16, intimacy information relating to the word or the like is added to and recorded on the individual text information recorded in units of the word or the like. This familiarity information is a numerical value (for example, 1.0 to 7.0) representing the word familiarity for each word or the like. Thus, the text analysis dictionary 16 constitutes the familiarity information storage means of the present invention.

  The language processing unit 15 refers to the text analysis dictionary 16 based on the text information input from the text generation unit 11, and obtains phoneme information and prosodic information corresponding to a character string such as a word indicated by the text information. By adding, the information of the phonetic character string is generated. At this time, the language processing unit 15 also reads the familiarity information stored corresponding to the input text information.

  The phoneme segment dictionary 18 is a phoneme segment dictionary database that stores waveform information to be added to character strings in units of character strings composed of various words. The speech synthesizer 17 refers to the phoneme segment dictionary 18 based on the information of the phonetic character string output from the language processing unit 15, and processes the phonetic character string using the waveform information, thereby synthesizing the synthesized speech. Is generated.

  The sound pressure adjustment unit 13 controls the volume 14 based on the familiarity information extracted from the text analysis dictionary 16 by the language processing unit 15, thereby adjusting the sound pressure level of the synthesized speech generated by the speech synthesis unit 17. Adjust in units of words. For example, the word having the highest word familiarity is set as the reference sound pressure level, and the volume value of the word having the lower word familiarity is corrected to be larger. The extent to which the volume value is increased is preferably different depending on the value of word familiarity as in the first embodiment.

  As described above in detail, according to the second embodiment, words and the like are compared with the text analysis dictionary 16 included in the TTS engine 12 that synthesizes and reproduces a speech waveform based on given text information. Since the familiarity information is added and stored in units, and the sound pressure level of each word to be output is appropriately adjusted based on the familiarity information extracted in the analysis of the text information, Even when outputting sounds that cannot be heard normally, it is possible to provide highly understandable (easy to hear) sounds. Therefore, for example, even when a navigation device to which the audio output device of the present embodiment is applied is used in an unfamiliar land, the guide audio is always easy to hear.

  In the second embodiment, the example in which the word having the highest word familiarity is set as the reference sound pressure level and the sound pressure level of the word having a lower word familiarity is greatly corrected has been described. The invention is not limited to this. For example, a word having the lowest word familiarity may be used as the reference sound pressure level, and a word having a medium word familiarity may be used as the reference sound pressure level.

  In the second embodiment as well, it is not always necessary to adjust the sound pressure level according to the word familiarity so that the intelligibility of all the words becomes equal, and the word understanding is achieved by adjusting the sound pressure. The degree may be larger than a predetermined value.

  In the second embodiment, the example in which the sound pressure level of the sound is adjusted based on the familiarity information has been described. However, in addition to or instead of this, a word whose word familiarity is lower than a predetermined value Etc. may be repeated two or more times. This repetitive playback control can be performed, for example, by the speech synthesizer 17 repeatedly synthesizing the same word or the like twice.

  Further, in addition to or instead of adjusting the sound pressure level of sound based on the familiarity information, the sound reproduction speed may be adjusted. This control of the playback speed can be performed, for example, by changing the output timing when the synthesized speech is output from the speech synthesizer 17.

  Further, in addition to or instead of adjusting the sound pressure level of the voice based on the familiarity information, a word having a word familiarity lower than a predetermined value may be displayed on the screen. This display control can be performed using a display controller (not shown) (for example, in the case of a navigation device, one that is normally provided for displaying a map image or the like on the display device).

(Third embodiment)
Next, a third embodiment according to the present invention will be described with reference to the drawings. In the third embodiment, the present invention is applied to a speech intelligibility improvement system using a loudness compensation technique. FIG. 3 is a diagram illustrating a configuration example of a main part of the speech intelligibility improvement system according to the third embodiment.

  As shown in FIG. 3, the speech intelligibility improving system of the present embodiment includes an audio DB 21, a playback unit 22, a volume or equalizer (hereinafter simply referred to as a volume or the like) 23, a sound pressure adjustment unit 24, a gain control unit 25, An adaptive filter (ADF) 26, a speaker 27, a microphone 28 and a subtractor 29 are provided.

  The audio DB 21 is configured by recording waveform-encoded audio data on a medium such as a DVD or a hard disk. In the voice DB 21, voice data to be output is recorded in units of words or the like. For example, when the speech intelligibility improving system of this embodiment is applied to a navigation device, a series of navigation voices such as “Following XX direction, traffic jam” is expressed as | Are recorded as separate audio patterns.

  The voice DB 21 further stores intimacy information added to individual voice patterns recorded in units of words or the like. This familiarity information is a numerical value (for example, 1.0 to 7.0) indicating how many word familiarities are in units of words or the like. Thus, the voice DB 21 constitutes the familiarity information storage means of the present invention.

  The reproduction unit 22 reproduces audio data and familiarity information from the audio DB 21. The reproduction unit 22 arbitrarily selects and reads out a plurality of sound patterns from the sound DB 21 (selects and reads a tag at a data position where a sound pattern corresponding to the content to be uttered is recorded), so It is possible to reproduce navigation voice. At this time, the familiarity information stored corresponding to the read voice patterns is also read.

  The volume etc. 23 controls the volume of the navigation voice reproduced by the reproduction unit 22. The speaker 27 outputs a navigation voice whose sound pressure is corrected by the volume 23 or the like. The microphone 28 is used for speech voice input. Actually, it is not only a voice command that is spoken, but also a navigation voice that is output from the speaker 27, an audio sound that is output from another speaker (not shown), and traveling noise. (Hereinafter, the audio sound and the running noise are collectively referred to as “ambient noise”) are also input to the same microphone 28.

  The adaptive filter 26 includes a coefficient identification unit and a voice correction filter. The coefficient identification unit is a filter for identifying a transfer function (a filter coefficient of a sound correction filter) of an acoustic system between the speaker 27 and the microphone 28, and is an LMS (Least Mean Square) algorithm or an N-LMS (Normalized-LMS). ) An adaptive filter based on an algorithm is used. The coefficient identification unit operates so as to minimize the power of an error signal (described later) output from the subtractor 29, and identifies the impulse response of the acoustic system.

  The sound correction filter performs a convolution operation using the filter coefficient determined by the coefficient identification unit and the sound pressure-corrected navigation sound to be controlled, thereby performing the above-described sound correction on the sound-corrected navigation sound. Gives the same transfer characteristics as the system. Thereby, a navigation simulation voice that simulates the navigation voice at the position of the microphone 28 is generated.

  The subtractor 29 extracts the peripheral noise by subtracting the navigation simulated voice generated by the adaptive filter 26 from the voice input from the microphone 28 (voice mixed with navigation voice and ambient noise). The ambient noise extracted by the subtractor 29 is fed back to the coefficient identification unit and the gain control unit 25 of the adaptive filter 26 as an error signal.

  Based on the simulated navigation sound output from the adaptive filter 26 and the ambient noise output from the subtractor 29, the gain control unit 25 is optimally added to the controlled navigation sound reproduced by the reproduction unit 22. The gain is calculated, and the calculated gain value is output to the sound pressure adjustment unit 24. Here, the surrounding noise (error signal) is regarded as noise for the navigation voice, and the gain of the navigation voice is adjusted so that the navigation voice output from the speaker 27 can be clearly heard by the user. Thus, the gain control unit 25 constitutes a gain calculation unit of the present invention.

  The sound pressure adjustment unit 24 controls the volume 23 based on the correction gain calculated by the gain control unit 25, adjusts the sound pressure level of the navigation sound to be output as a whole, and the sound DB 21 by the playback unit 22. The volume or the like 23 is controlled based on the familiarity information read out from, and the sound pressure level of the navigation voice to be output is adjusted in units of words or the like. For example, the word having the highest word familiarity is set as the reference sound pressure level, and the volume value of the word having the lower word familiarity is corrected to be larger.

  For example, when outputting a navigation voice such as “This is the direction of XX ahead, traffic jam”, the volume value is adjusted as a whole for a series of navigation voices so that this navigation voice can be heard clearly even if there is a surrounding sound. To do. Furthermore, if the part of XX that represents the place name is a word with a low word familiarity such as “Daisen” or “Daizuki”, add more compensation for that word section Adjust the volume value. As with the first embodiment, it is preferable that how much the volume value is increased in units of words or the like depends on the value of word familiarity.

  As described above in detail, according to the third embodiment, in the loudness compensation type speech intelligibility improvement system, the speech compensation amount is appropriately adjusted in units of words based on the word familiarity information. It is possible to make the output target sound clearly audible regardless of the ambient sound, and to make it easy to hear even when outputting a sound whose contents cannot be heard normally. Therefore, for example, even when a navigation device to which this speech intelligibility improvement system is applied is used in an unfamiliar land, the guide speech is always easy to hear.

  In the third embodiment, the example in which the word having the highest word familiarity is set as the reference sound pressure level and the sound pressure level of the word having a lower word familiarity is corrected to a greater value has been described. The invention is not limited to this. For example, a word having the lowest word familiarity may be used as the reference sound pressure level, and a word having a medium word familiarity may be used as the reference sound pressure level.

  In the third embodiment as well, it is not always necessary to adjust the sound pressure level in accordance with the word familiarity so that the intelligibility of all the words becomes the same level. The degree may be larger than a predetermined value.

  In the third embodiment, the example in which the sound pressure level of the sound is adjusted based on the familiarity information has been described, but in addition to or instead of this, a word whose word familiarity is lower than a predetermined value Etc. may be repeated two or more times. This repetitive playback control can be performed by the playback unit 22.

  Further, in addition to or instead of adjusting the sound pressure level of sound based on the familiarity information, the sound reproduction speed may be adjusted. The playback speed can also be controlled by the playback unit 22.

  Further, in addition to or instead of adjusting the sound pressure level of the voice based on the familiarity information, a word having a word familiarity lower than a predetermined value may be displayed on the screen. This display control can be performed using a display controller (not shown) (for example, in the case of a navigation device, one that is normally provided for displaying a map image or the like on the display device).

(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described with reference to the drawings. In the fourth embodiment, the present invention is applied to a voice call system (for example, a hands-free system). FIG. 4 is a diagram illustrating a configuration example of a main part of a voice call system according to the fourth embodiment.

  As shown in FIG. 4, the voice call system of this embodiment includes an acoustic model DB 31, a language model DB 32, a first continuous recognition unit 33, a first sound pressure adjustment unit 34, a first volume 35, a speaker 36, A microphone 37, a second continuous recognition unit 38, a second sound pressure adjustment unit 39, and a second volume 40 are provided.

  The acoustic model DB 31 is a speech dictionary database that stores a character string such as each word to be recognized and a feature amount of the speech pattern in association with each other. The language model DB 32 is a syntax analysis dictionary database that stores information necessary for analyzing the syntax of a recognized speech pattern. The language model DB 32 further stores information indicating the relationship between text information representing character strings such as various words and their closeness. As described above, the language model DB 32 constitutes the familiarity information storage unit of the present invention.

  The first continuous recognition unit 33 calculates the feature amount from the received voice, compares the calculated feature amount with the feature amount of each word or the like stored in advance in the acoustic model DB 31, and has the highest similarity. A voice pattern is searched, and a character string having the voice pattern is recognized as a character string of the received voice. Then, the input received voice is converted into text information of the recognized character string. Thus, the 1st continuous recognition part 33 comprises the 1st audio | voice recognition means of this invention.

  The first continuous recognition unit 33 refers to the language model DB 32 based on the converted text information, reads the familiarity information stored corresponding to the text information, and the first sound pressure adjustment unit 34. The processing to supply to is also performed. The first sound pressure adjusting unit 34 controls the first volume 35 based on the familiarity information supplied from the first continuous recognizing unit 33, thereby setting the sound pressure level of the received voice in units such as words. adjust. For example, the word having the highest word familiarity is set as the reference sound pressure level, and the volume value of the word having the lower word familiarity is corrected to be larger. The received voice with the corrected sound pressure is output from the speaker 36.

  The second continuous recognition unit 38 calculates a feature value from the transmitted voice input from the microphone 37, and compares the calculated feature value with a feature value of each word or the like stored in advance in the acoustic model DB 31. Then, the voice pattern having the highest similarity is searched, and the character string having the voice pattern is recognized as the character string of the transmitted voice. Then, the input transmission voice is converted into text information of the recognized character string. Thus, the 2nd continuous recognition part 38 comprises the 2nd audio | voice recognition means of this invention.

  The second continuous recognition unit 38 refers to the language model DB 32 based on the converted text information, reads the familiarity information stored corresponding to the text information, and the second sound pressure adjustment unit 39. The processing to supply to is also performed. The second sound pressure adjustment unit 39 controls the second volume 40 based on the familiarity information supplied from the second continuous recognition unit 38, thereby changing the sound pressure level of the transmitted voice to a unit such as a word. Adjust with. For example, the word having the highest word familiarity is set as the reference sound pressure level, and the volume value of the word having the lower word familiarity is corrected to be larger. The transmitted voice with the corrected sound pressure is transmitted to the other party.

  In the example of FIG. 4, a continuous recognition unit and a sound pressure adjustment unit are provided on both the reception side and the transmission side. As a result, even if the other party's voice communication system does not have the same configuration as that shown in FIG. 4, it is possible to provide call voice with sound pressure adjusted appropriately according to the utterance contents for both transmission and reception. In the present invention, it is not always necessary to provide the continuous recognition unit and the sound pressure adjustment unit on both the reception side and the transmission side, and only one of them may be provided.

  In addition, when the continuous recognition unit and the sound pressure adjustment unit are provided on both the receiver side and the transmitter side, if the other side also has the same configuration, the sound whose sound pressure has been adjusted on its own transmitter side The sound pressure is further adjusted on the other party's receiver side, and the sound pressure is adjusted more than necessary. Therefore, before starting a call (at the time of the first call), it is confirmed whether or not the other party has a sound pressure adjustment unit by performing predetermined communication. When the other party has a sound pressure adjustment unit, it is possible to control so that at least one of the functions of the first sound pressure adjustment unit 34 and the second sound pressure adjustment unit 39 is suspended. is there.

  For example, when making a call for the first time, an inquiry signal is transmitted from the calling-side voice communication system to the called-side voice communication system to inquire whether or not the called-side has a sound pressure adjustment unit. In response to this inquiry, the called side system returns the presence or absence of the sound pressure adjustment unit to the calling side system. When receiving a reply that there is a sound pressure adjusting unit, the calling side system controls to stop the function of the first sound pressure adjusting unit 34 in the calling side system. In addition, a signal instructing to suspend the function of the first sound pressure adjusting unit 34 in the called side system is transmitted to the called side system, and the function of the first sound pressure adjusting unit 34 in the called side system is also provided. Control to pause.

  When the calling side system receives a response from the called side system that there is a sound pressure adjusting unit, the second sound pressure adjusting unit 39 in the calling side system and the second sound in the called side system. You may control to make the function with the pressure adjustment part 39 pause. Alternatively, the functions of the first sound pressure adjustment unit 34 and the second sound pressure adjustment unit 39 in the calling side system may be controlled to be suspended, and the functions of the called side system may not be suspended. Furthermore, the functions of the sound pressure adjustment units 34 and 39 may not be suspended in both the calling side and called side systems, and the increase / decrease width of the sound pressure may be controlled to be about half of the normal.

  As described above in detail, according to the fourth embodiment, in the voice communication system, the call voice is recognized and parsed, and the sound pressure in units of words or the like based on the word familiarity information using the analysis result. Therefore, even if there is an unfamiliar utterance content during a call, it can be easily heard by correcting the sound pressure. Therefore, a comfortable call can always be made.

  In the fourth embodiment, the example in which the word having the highest word familiarity is set as the reference sound pressure level and the sound pressure level of the word having a lower word familiarity is corrected to a greater value has been described. The invention is not limited to this. For example, a word having the lowest word familiarity may be used as the reference sound pressure level, and a word having a medium word familiarity may be used as the reference sound pressure level.

  In the fourth embodiment, the word intelligibility may be made larger than a predetermined value by adjusting the sound pressure, and the intelligibility of all the words and the like is not necessarily equal. .

  In the fourth embodiment, the example in which the sound pressure level of the sound is adjusted based on the familiarity information has been described, but in addition to or instead of this, the word whose word familiarity is lower than a predetermined value Etc. may be repeated two or more times. This repeated reproduction control can be performed as follows, for example. That is, the received voice and the transmitted voice are digitized and temporarily stored in the buffer memory, and the reading from the buffer memory is repeated twice or more, and the read voice is returned to the analog signal again.

  Further, in addition to or instead of adjusting the sound pressure level of the sound based on the familiarity information, the sound reproduction speed may be adjusted according to the word familiarity. This reproduction speed control can also be performed, for example, as follows. That is, the received voice and the transmitted voice are digitized and temporarily stored in the buffer memory, and the read timing from the buffer memory is made variable according to the word familiarity.

  Further, in addition to or instead of adjusting the sound pressure level of the voice based on the familiarity information, a word having a word familiarity lower than a predetermined value may be displayed on the screen. This display control can be performed using a display controller (not shown) (for example, one that is usually provided for displaying a telephone number or the like on the display device).

  The sound pressure adjustment methods according to the first to fourth embodiments described above can be realized by any of a hardware configuration, a DSP, and software. For example, when realized by software, the audio output device of the present embodiment is actually configured by including a CPU or MPU of a computer, RAM, ROM, and the like, and can be realized by operating a program stored in the RAM or ROM. .

  Therefore, it can be realized by recording a program that causes a computer to perform the functions of the above-described embodiments on a recording medium such as a CD-ROM and causing the computer to read the program. As a recording medium for recording the program, a flexible disk, a hard disk, a magnetic tape, an optical disk, a magneto-optical disk, a DVD, a nonvolatile memory card, and the like can be used in addition to the CD-ROM. It can also be realized by downloading the program to a computer via a network such as the Internet.

  The above first to fourth embodiments are merely examples of implementation in carrying out the present invention, and the technical scope of the present invention should not be construed as being limited thereto. It will not be. In other words, the present invention can be implemented in various forms without departing from the spirit or main features thereof.

  The voice output device and method of the present invention can be widely applied to a device or system that adjusts sound pressure in units of words or the like according to familiarity. For example, the navigation device and voice call described in the above embodiment Useful for systems. The present invention is also useful for an information communication apparatus having a function of reading out the contents of Web information and electronic mail received from an information network such as the Internet. Furthermore, in a language learning system that reads out conversations, words, and the like, it is also useful when adjusting the sound pressure of words that are difficult to pronounce, such as words that cannot be heard or words that have a high degree of difficulty.

It is a figure which shows the principal part structural example of the audio | voice output apparatus which concerns on 1st Embodiment. It is a figure which shows the principal part structural example of the audio | voice output apparatus which concerns on 2nd Embodiment. It is a figure which shows the principal part structural example of the speech intelligibility improvement system which concerns on 3rd Embodiment. It is a figure which shows the principal part structural example of the voice call system which concerns on 4th Embodiment. It is a characteristic view of the test result which showed the relationship between word familiarity, word intelligibility, and sound pressure.

Explanation of symbols

1 Voice DB
2 playback unit 3 sound pressure adjustment unit 4 volume 11 text generation unit 12 TTS engine 13 sound pressure adjustment unit 14 volume 15 language processing unit 16 text analysis dictionary 17 speech synthesis unit 18 phoneme segment dictionary 21 speech DB
22 playback unit 23 volume or equalizer 24 sound pressure adjustment unit 25 gain control unit 26 adaptive filter 27 speaker 28 microphone 29 subtractor 31 acoustic model DB
32 Language Model DB
33 1st continuous recognition part 34 1st sound pressure adjustment part 35 1st volume 36 speaker 37 microphone 38 2nd continuous recognition part 39 2nd sound pressure adjustment part 40 2nd volume

Claims (20)

Intimacy information storage means for storing information about intimacy indicating how familiar a plurality of words or word strings are;
A sound output device comprising: sound pressure adjustment means for adjusting the sound pressure level of each word or word string to be output based on the familiarity information stored in the familiarity information storage means. The intimacy information storage means is configured by adding the intimacy information in units of words or word strings to an audio database in which each word or word string to be output is recorded. The audio output device according to claim 1, wherein The closeness information storage means stores the closeness information in units of words or word strings with respect to a dictionary database for text analysis provided in a device that synthesizes and reproduces a speech waveform based on given text information. The audio output device according to claim 1, wherein the audio output device is configured to be added. Gain calculating means for calculating a correction gain of the output sound based on the sound pressure level of the output sound and the sound pressure level of ambient sound that can be heard at the listening position of the output sound,
The sound pressure adjusting means adjusts the sound pressure level of the sound to be output based on the correction gain calculated by the gain calculating means, and adds the familiarity information stored in the familiarity information storage means. The sound output device according to claim 1, wherein the sound pressure level of the sound to be output is adjusted based on a word or a word string. Comparing the input voice with a prepared voice dictionary, the voice recognition means for recognizing a word or word string related to the input voice and converting it into text information,
The intimacy information storage means is configured to store information indicating a relationship between text information representing a plurality of words or word strings and the intimacy,
The sound pressure adjusting means determines the sound pressure level of the input speech based on the familiarity information obtained by referring to the familiarity information storage means based on the text information converted by the speech recognition means. The audio output device according to claim 1, wherein adjustment is performed in units of word strings. The voice recognition means is characterized by inputting a received voice in a voice call system, comparing with a voice dictionary prepared in advance, recognizing a word or word string related to the received voice and converting it into text information. The audio output device according to claim 5. The voice recognition means inputs a voice to be spoken in a voice call system, compares it with a voice dictionary prepared in advance, recognizes a word or word string related to the voice to be sent, and converts it into text information. The audio output device according to claim 5, wherein: The voice recognition means inputs a received voice in a voice call system, compares it with a voice dictionary prepared in advance, recognizes a word or a word string related to the received voice and converts it into text information. A recognition unit is compared with a voice dictionary prepared in advance by inputting a transmission voice in the voice call system, and a second word or word string related to the transmission voice is recognized and converted into text information. Voice recognition means,
The sound pressure adjusting means is configured to generate a sound pressure of the received voice based on the familiarity information obtained by referring to the familiarity information storage means based on the text information converted by the first speech recognition means. The first sound pressure adjusting means for adjusting the level in units of words or word strings, and the familiarity information storing means obtained by referring to the familiarity information storage means based on the text information converted by the second speech recognition means 6. The voice output device according to claim 5, further comprising second sound pressure adjusting means for adjusting a sound pressure level of the transmitted voice in units of words or word strings based on familiarity information. Before starting a call in the voice call system, determining means for determining whether the other party is equipped with the sound pressure adjusting means;
When it is determined by the determining means that the other party has the sound pressure adjusting means, at least one function of the first sound pressure adjusting means and the second sound pressure adjusting means is suspended. 9. The audio output device according to claim 8, further comprising a control unit for controlling. A reproduction control means for controlling to reproduce a word or a word string having a familiarity lower than a predetermined value twice or more based on the familiarity information stored in the familiarity information storage means. The audio output device according to claim 1. The reproduction control means for adjusting the reproduction speed of each word or word string to be output based on the familiarity information stored in the familiarity information storage means. Audio output device. A display control means for controlling to display on the screen a word or a word string whose intimacy is lower than a predetermined value based on the intimacy information stored in the intimacy information storage means. The audio output device according to claim 1. The sound pressure adjustment unit refers to the familiarity information that indicates how familiar the multiple words or word strings are, and the sound pressure level of each word or word string to be output depends on the familiarity information. The audio output method is characterized in that the adjustment is performed. When the sound is reproduced from the sound database in which each word or word string to be output is recorded, the sound pressure adjusting unit stores the intimacy information recorded in units of words or word strings on the sound database. 14. The audio output method according to claim 13, wherein the sound pressure level of the audio to be reproduced is adjusted in units of words or word strings. The sound pressure adjustment unit refers to the intimacy information recorded in units of words or word strings in the dictionary database for text analysis when synthesizing and reproducing a speech waveform based on the given text information The sound output method according to claim 13, wherein the sound pressure level of the sound to be reproduced is adjusted in units of words or word strings. When playing back speech input from the outside, the speech recognition unit recognizes the word or word string related to the input speech by comparing the input speech with a speech dictionary prepared in advance, and the recognized word 14. The voice according to claim 13, wherein the sound pressure adjustment unit refers to the familiarity information corresponding to a word string, and adjusts the sound pressure level of the reproduced voice in units of words or word strings. output method. A correction gain is obtained based on the sound pressure level of the output sound and the sound pressure level of the ambient sound that can be heard at the listening position of the output sound, and the sound pressure level of the output sound is corrected based on the correction gain. In the speech intelligibility improvement system
The sound pressure adjusting unit adjusts the sound pressure level of the output sound based on the correction gain, and adjusts the sound pressure level of the output sound in units of words or word strings based on the familiarity information. The audio output method according to claim 13, wherein: 14. The audio output method according to claim 13, wherein a word or a word string having a familiarity lower than a predetermined value is repeatedly reproduced and output based on the familiarity information twice or more. Based on the familiarity information, a word or a word string having a familiarity lower than a predetermined value is reproduced and output at a slower speed than a word or a word string having a familiarity greater than or equal to the predetermined value. The audio output method according to claim 13. 14. The audio output method according to claim 13, wherein a word or a word string whose intimacy is lower than a predetermined value is displayed on the screen based on the intimacy information.

Images